Some say that the technical platform future lies with commodity-based blade computers.
Yet others say that software as a service will drive us towards a socket-in-the-wall mentality where all computer requirements are served by hosted managed solutions.
Are any of these right? On their own, probably not. What is more likely is that they will all be right in one way or another.
As an analogy, I would like to use the history of our planet, and how the mountains and plains have emerged over time.
Right back at the beginning, we had a spinning ball of debris with little topographical detail - which we can compare to the pre-computer days of abacuses, tally sticks, log tables and slide rules, but nothing very big.
The planet then consolidated and formed a hard crust, with contents that over a period of billions of years crashed into each other, creating huge mountain chains (for us, mainframes).
On these mountains, luxurious flora flourished, much like the specialised applications written by the early developers who had no need to worry about standards and interoperability.
Away from the edges of the main activity, large areas of relative stability remained as plains (the majority of companies, which did not use computers at all).
Even today, ongoing volcanic and tectonic plate activity still sends up a new mountain every now and again.
The early mountains wore slowly down to smaller mountains (midi servers) and rolling hills and dales (distributed client/server computing).
Water and ice action created valleys even in the plains (even laggard organisations eventually found computers useful).
Man enters the picture very late on, and, finding the valleys fertile and easy to farm, slowly moves over to common crops such as wheat and barley - in much the same way as early standardisation happened in the client/server world.
It gives us today's "standard" environment - flood plains, valleys, foothills and mountains.
To make things work for us, we have globalised agriculture in areas with a similar environment (the right sort of soil, weather and farming capabilities), more specialised approaches for areas with less commonality, and exotica growing in the higher reaches where few people dare to tread.
But erosion will continue. The mountains and many of the larger hills are ill-suited to long-term survival.
For example, large rocks are precariously perched on steep slopes, cliffs are broken down by ice and water, thin layers of once heavily forested soil are washed away.
Similarly, within mainframe and large server applications, redundant functionality will be washed out into the valleys and flood plains, which will silt up with smaller, commodity blade computers.
But the large machines will still be there, doing what they are best at, being more "exotic" than the heavily farmed low-lying areas.
How we access the world has also changed. In the early days, climbing a mountain was something that a few could be proud of, but now even if you cannot scale a mountain on foot, you can fly around it and see its glories up close.
With the technology world, we have driven connectivity and interoperability, opening up the large mountains and the rolling hills to the flood plain dwellers.
The next generation of service-oriented composite applications will find a way up to the highest parts of the mountains.
For the purist, this has little attraction: being one of the few who can physically go there is a big part of the attraction.
But we may be able to keep everyone happy - for example by virtualising the whole experience and allowing people to "be there" from the real comfort of their own environment.
Similarly in IT, there will always be those who prefer to keep all others out of their domain. But integration and the removal of functional redundancy is all to the good.
And the ending to this analogy? It is that you cannot change what you have got completely, and the world will only change slowly around you.
But in a world in which you are just a very small item, changing your viewpoint can let you appreciate things a lot more.
You can download the PDF version of this article here